support G711 tracks with multiple channels and custom sample rates (#2891)

go.mod

@@ -8,7 +8,7 @@ require (
 	github.com/alecthomas/kong v0.8.1
 	github.com/aler9/writerseeker v1.1.0
 	github.com/bluenviron/gohlslib v1.2.0
-	github.com/bluenviron/gortsplib/v4 v4.6.3
+	github.com/bluenviron/gortsplib/v4 v4.6.4-0.20240108201647-63a81d0896f5
 	github.com/bluenviron/mediacommon v1.7.1
 	github.com/datarhei/gosrt v0.5.5
 	github.com/fsnotify/fsnotify v1.7.0

go.sum

@@ -22,8 +22,8 @@ github.com/benburkert/openpgp v0.0.0-20160410205803-c2471f86866c h1:8XZeJrs4+ZYh
 github.com/benburkert/openpgp v0.0.0-20160410205803-c2471f86866c/go.mod h1:x1vxHcL/9AVzuk5HOloOEPrtJY0MaalYr78afXZ+pWI=
 github.com/bluenviron/gohlslib v1.2.0 h1:Hrx2/n/AcmKKIV+MjZLKs5kmW+O7xCdUSPJQoS39JKw=
 github.com/bluenviron/gohlslib v1.2.0/go.mod h1:kG/Sjebsxnf5asMGaGcQ0aSvtFGNChJPgctds2wDHOI=
-github.com/bluenviron/gortsplib/v4 v4.6.3 h1:/FgUhxe/DGFkmwHNGNnw7NhcuZeSwnxCzYHDbyAtMjo=
+github.com/bluenviron/gortsplib/v4 v4.6.4-0.20240108201647-63a81d0896f5 h1:DU63YvVivfcIq7jcA1KCwi9YO+jEVglYNJCIrOLj8QE=
-github.com/bluenviron/gortsplib/v4 v4.6.3/go.mod h1:UqdkRR5YvKHP/wHEQQySJFKJm6tIZcftdP7cNszIZ1g=
+github.com/bluenviron/gortsplib/v4 v4.6.4-0.20240108201647-63a81d0896f5/go.mod h1:UqdkRR5YvKHP/wHEQQySJFKJm6tIZcftdP7cNszIZ1g=
 github.com/bluenviron/mediacommon v1.7.1 h1:7Lm2b8M9gUk3Ben1w+OPwadAeYseIBscwgdiL+aLtfw=
 github.com/bluenviron/mediacommon v1.7.1/go.mod h1:Ij/kE1LEucSjryNBVTyPL/gBI0d6/Css3f5PyrM957w=
 github.com/bytedance/sonic v1.5.0/go.mod h1:ED5hyg4y6t3/9Ku1R6dU/4KyJ48DZ4jPhfY1O2AihPM=

internal/formatprocessor/g711.go

@@ -5,7 +5,7 @@ import (
 	"time"
 
 	"github.com/bluenviron/gortsplib/v4/pkg/format"
-	"github.com/bluenviron/gortsplib/v4/pkg/format/rtpsimpleaudio"
+	"github.com/bluenviron/gortsplib/v4/pkg/format/rtplpcm"
 	"github.com/pion/rtp"
 
 	"github.com/bluenviron/mediamtx/internal/unit"
@@ -14,8 +14,8 @@ import (
 type formatProcessorG711 struct {
 	udpMaxPayloadSize int
 	format            *format.G711
-	encoder           *rtpsimpleaudio.Encoder
+	encoder           *rtplpcm.Encoder
-	decoder           *rtpsimpleaudio.Decoder
+	decoder           *rtplpcm.Decoder
 }
 
 func newG711(
@@ -39,9 +39,11 @@ func newG711(
 }
 
 func (t *formatProcessorG711) createEncoder() error {
-	t.encoder = &rtpsimpleaudio.Encoder{
+	t.encoder = &rtplpcm.Encoder{
 		PayloadMaxSize: t.udpMaxPayloadSize - 12,
 		PayloadType:    t.format.PayloadType(),
+		BitDepth:       8,
+		ChannelCount:   t.format.ChannelCount,
 	}
 	return t.encoder.Init()
 }
@@ -49,14 +51,16 @@ func (t *formatProcessorG711) createEncoder() error {
 func (t *formatProcessorG711) ProcessUnit(uu unit.Unit) error { //nolint:dupl
 	u := uu.(*unit.G711)
 
-	pkt, err := t.encoder.Encode(u.Samples)
+	pkts, err := t.encoder.Encode(u.Samples)
 	if err != nil {
 		return err
 	}
-	u.RTPPackets = []*rtp.Packet{pkt}
+	u.RTPPackets = pkts
 
 	ts := uint32(multiplyAndDivide(u.PTS, time.Duration(t.format.ClockRate()), time.Second))
-	u.RTPPackets[0].Timestamp += ts
+	for _, pkt := range u.RTPPackets {
+		pkt.Timestamp += ts
+	}
 
 	return nil
 }

internal/formatprocessor/g711_test.go

@@ -12,7 +12,10 @@ import (
 func TestG611Encode(t *testing.T) {
 	t.Run("alaw", func(t *testing.T) {
 		forma := &format.G711{
-			MULaw: false,
+			PayloadTyp:   8,
+			MULaw:        false,
+			SampleRate:   8000,
+			ChannelCount: 1,
 		}
 
 		p, err := New(1472, forma, true)
@@ -37,7 +40,10 @@ func TestG611Encode(t *testing.T) {
 
 	t.Run("mulaw", func(t *testing.T) {
 		forma := &format.G711{
-			MULaw: true,
+			PayloadTyp:   0,
+			MULaw:        true,
+			SampleRate:   8000,
+			ChannelCount: 1,
 		}
 
 		p, err := New(1472, forma, true)

internal/protocols/rtmp/reader.go

@@ -95,17 +95,9 @@ func hasAudio(md flvio.AMFMap, audioTrack *format.Format) (bool, error) {
 			return true, nil
 
 		case message.CodecPCMA:
-			v, ok := md.GetV("stereo")
-			if ok && v == true {
-				return false, fmt.Errorf("stereo PCMA is not supported")
-			}
 			return true, nil
 
 		case message.CodecPCMU:
-			v, ok := md.GetV("stereo")
-			if ok && v == true {
-				return false, fmt.Errorf("stereo PCMU is not supported")
-			}
 			return true, nil
 		}
 
@@ -336,19 +328,31 @@ func tracksFromMetadata(conn *Conn, payload []interface{}) (format.Format, forma
 					}
 
 				case msg.Codec == message.CodecPCMA:
-					if msg.Channels == message.ChannelsStereo {
+					audioTrack = &format.G711{
-						return nil, nil, fmt.Errorf("stereo PCMA is not supported")
+						PayloadTyp: 8,
+						MULaw:      false,
+						SampleRate: 8000,
+						ChannelCount: func() int {
+							if msg.Channels == message.ChannelsStereo {
+								return 2
+							}
+							return 1
+						}(),
 					}
 
-					audioTrack = &format.G711{MULaw: false}
-
 				case msg.Codec == message.CodecPCMU:
-					if msg.Channels == message.ChannelsStereo {
+					audioTrack = &format.G711{
-						return nil, nil, fmt.Errorf("stereo PCMU is not supported")
+						PayloadTyp: 0,
+						MULaw:      true,
+						SampleRate: 8000,
+						ChannelCount: func() int {
+							if msg.Channels == message.ChannelsStereo {
+								return 2
+							}
+							return 1
+						}(),
 					}
 
-					audioTrack = &format.G711{MULaw: true}
-
 				case msg.Codec == message.CodecLPCM:
 					audioTrack = &format.LPCM{
 						PayloadTyp: 96,

internal/protocols/rtmp/reader_test.go

@@ -786,7 +786,12 @@ func TestReadTracks(t *testing.T) {
 		{
 			"pcma",
 			nil,
-			&format.G711{},
+			&format.G711{
+				PayloadTyp:   8,
+				MULaw:        false,
+				SampleRate:   8000,
+				ChannelCount: 1,
+			},
 			[]message.Message{
 				&message.DataAMF0{
 					ChunkStreamID:   4,
@@ -817,7 +822,10 @@ func TestReadTracks(t *testing.T) {
 			"pcmu",
 			nil,
 			&format.G711{
-				MULaw: true,
+				PayloadTyp:   0,
+				MULaw:        true,
+				SampleRate:   8000,
+				ChannelCount: 1,
 			},
 			[]message.Message{
 				&message.DataAMF0{

internal/protocols/webrtc/incoming_track.go

@@ -80,12 +80,18 @@ func newIncomingTrack(
 
 	case strings.ToLower(webrtc.MimeTypePCMU):
 		t.format = &format.G711{
-			MULaw: true,
+			PayloadTyp:   0,
+			MULaw:        true,
+			SampleRate:   8000,
+			ChannelCount: 1,
 		}
 
 	case strings.ToLower(webrtc.MimeTypePCMA):
 		t.format = &format.G711{
-			MULaw: false,
+			PayloadTyp:   8,
+			MULaw:        false,
+			SampleRate:   8000,
+			ChannelCount: 1,
 		}
 
 	default:

internal/record/agent_test.go

@@ -56,13 +56,19 @@ func TestAgent(t *testing.T) {
 		{
 			Type: description.MediaTypeAudio,
 			Formats: []rtspformat.Format{&rtspformat.G711{
-				MULaw: false,
+				PayloadTyp:   8,
+				MULaw:        false,
+				SampleRate:   8000,
+				ChannelCount: 1,
 			}},
 		},
 		{
 			Type: description.MediaTypeAudio,
 			Formats: []rtspformat.Format{&rtspformat.G711{
-				MULaw: true,
+				PayloadTyp:   0,
+				MULaw:        true,
+				SampleRate:   8000,
+				ChannelCount: 1,
 			}},
 		},
 	}}

internal/record/format_fmp4.go

@@ -775,8 +775,8 @@ func (f *formatFMP4) initialize() {
 				codec := &fmp4.CodecLPCM{
 					LittleEndian: false,
 					BitDepth:     16,
-					SampleRate:   8000,
+					SampleRate:   forma.SampleRate,
-					ChannelCount: 1,
+					ChannelCount: forma.ChannelCount,
 				}
 				track := addTrack(codec)
 

internal/servers/webrtc/session.go

@@ -244,6 +244,14 @@ func findAudioTrack(
 
 	if g711Format != nil {
 		return g711Format, func(track *webrtc.OutgoingTrack) error {
+			if g711Format.SampleRate != 8000 {
+				return fmt.Errorf("unsupported G711 sample rate")
+			}
+
+			if g711Format.ChannelCount != 1 {
+				return fmt.Errorf("unsupported G711 channel count")
+			}
+
 			stream.AddReader(writer, media, g711Format, func(u unit.Unit) error {
 				for _, pkt := range u.GetRTPPackets() {
 					track.WriteRTP(pkt) //nolint:errcheck